BitSage MCP Server
Enables LLMs to interact with the BitSage Network for distributed compute operations including job management, worker discovery, proof verification, staking, and network monitoring.
README
BitSage MCP Server
MCP (Model Context Protocol) server that enables LLMs like Claude to interact with the BitSage Network for distributed compute operations.
Features
- Job Management: Submit, monitor, and cancel compute jobs
- Worker Discovery: List and inspect available compute workers
- Proof Verification: Verify ZK proofs on-chain
- Staking: Stake SAGE tokens and claim rewards
- Faucet: Claim testnet tokens (Sepolia only)
- Network Stats: Monitor network health and utilization
Installation
npm install @bitsage/mcp-server
Or run directly with npx:
npx @bitsage/mcp-server
Configuration
Set environment variables:
# API endpoint (optional, defaults to production)
export BITSAGE_API_URL=https://api.bitsage.network
# Starknet RPC (optional)
export STARKNET_RPC_URL=https://starknet-sepolia.public.blastapi.io
# Network: mainnet or sepolia (optional, defaults to sepolia)
export BITSAGE_NETWORK=sepolia
Usage with Claude Desktop
Add to your Claude Desktop configuration (~/.claude/claude_desktop_config.json):
{
"mcpServers": {
"bitsage": {
"command": "npx",
"args": ["@bitsage/mcp-server"],
"env": {
"BITSAGE_NETWORK": "sepolia"
}
}
}
}
Available Tools
Job Management
| Tool | Description |
|---|---|
bitsage_submit_job |
Submit a compute job (AI inference, ZK proof, etc.) |
bitsage_get_job_status |
Get current status of a job |
bitsage_cancel_job |
Cancel a pending or running job |
bitsage_list_jobs |
List jobs with optional status filter |
Worker Management
| Tool | Description |
|---|---|
bitsage_list_workers |
List available compute workers |
bitsage_get_worker |
Get details about a specific worker |
Proof Verification
| Tool | Description |
|---|---|
bitsage_get_proof |
Get proof details by hash |
bitsage_verify_proof |
Verify a proof on-chain |
Staking
| Tool | Description |
|---|---|
bitsage_stake |
Stake SAGE tokens |
bitsage_unstake |
Unstake SAGE tokens |
bitsage_claim_rewards |
Claim staking rewards |
bitsage_get_stake_info |
Get staking info for an address |
Faucet (Testnet only)
| Tool | Description |
|---|---|
bitsage_faucet_claim |
Claim testnet SAGE tokens |
bitsage_faucet_status |
Check faucet cooldown status |
ZKML Proving & Verification
| Tool | Description |
|---|---|
bitsage_submit_zkml_proof |
Submit a ZKML proving job (model must be loaded on prover) |
bitsage_get_zkml_proof_status |
Poll proving job status and progress |
bitsage_get_zkml_proof_result |
Get completed proof (calldata, commitments, gas estimate) |
bitsage_verify_zkml_onchain |
Check on-chain verification status for a proof or model |
Network
| Tool | Description |
|---|---|
bitsage_network_stats |
Get network statistics |
ZKML Configuration
The ZKML tools connect to a prove-server instance (see libs/stwo-ml/) and the on-chain verifier contract.
# Prover server URL (default: http://localhost:8080)
export BITSAGE_PROVER_URL=http://your-gpu-server:8080
# On-chain verifier contract (default: deployed v3 on Sepolia)
export ZKML_VERIFIER_ADDRESS=0x048070fbd531a0192f3d4a37eb019ae3174600cae15e08c737982fae5d929160
Example Interactions
Submit an AI Inference Job
User: Submit an AI inference job using llama-7b model with my prompt "What is the capital of France?"
Claude: I'll submit that job for you.
[Uses bitsage_submit_job tool]
The job has been submitted successfully:
- Job ID: abc123-def456
- Status: pending
- Estimated cost: 50 SAGE tokens
Check Network Status
User: How many workers are available on the BitSage network?
Claude: Let me check the network stats.
[Uses bitsage_network_stats tool]
The BitSage network currently has:
- Total workers: 150
- Active workers: 89
- Worker utilization: 59.3%
- Jobs in progress: 42
- Total jobs completed: 15,234
Prove and Verify ML Inference
User: Prove the model 0xabc123 and check if it verifies on-chain.
Claude: I'll submit a proving job and track it.
[Uses bitsage_submit_zkml_proof with model_id: "0xabc123", gpu: true]
Proving job submitted:
- Job ID: d4e5f6a7-...
- Status: queued
[Uses bitsage_get_zkml_proof_status with job_id: "d4e5f6a7-..."]
Proof completed in 40.5s:
- 160 matmul sumcheck proofs
- 40 layers proven
- ~350,000 estimated gas
- 2,847 calldata felts
[Uses bitsage_verify_zkml_onchain with model_id: "0xabc123"]
On-chain status:
- Verification count: 3
- Weight commitment: 0x7f2a...
Development
# Install dependencies
npm install
# Build
npm run build
# Run in development
npm run dev
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。